Quantifying the reproducibility of scientometric analyses: a case study